robertbarclayy/conducty/conducty-obsidian
Claude Code and agent skill by robertbarclayy/conducty
- community
Checking that the SKILL.md file is still in the repository.
Conducty's context engine lives in an Obsidian vault. This skill defines the vault layout, naming, linking, and index conventions every other Conducty skill must follow when reading or writing notes. Use when reading/writing any plan, design, context, improvement, failure pattern, metrics, or prompt-log note — or when the user says "vault", "Obsidian", "context engine", "where is X stored".
Install
npx vskill@latest install robertbarclayy/conducty/conducty-obsidianSKILL.md
Conducty Obsidian — The Context Engine
Conducty's plans, designs, context, improvements, failure patterns, metrics, and prompt logs all live in an Obsidian vault. The vault is not a backup or export target — it IS the context engine. Every Conducty note is written there, linked to its peers, and re-read on future sessions to inform planning.
[!important] Read this firstBefore writing or reading any state file (plan, design, context, etc.), check this skill for naming and linking conventions. Other Conducty skills assume you've internalized these rules.
Vault Location
Resolve the vault root in this order:
$CONDUCTY_VAULTenvironment variable, if set~/Obsidian/Conducty/fallback
When this skill says "the vault" or `{vault}/...`, substitute the resolved path. Always use absolute paths when invoking Bash/Read/Write/Edit.
If the vault directory doesn't exist, create it (and the seed index notes — see Bootstrap) before writing anything.
Why a Vault, Not Folders
Earlier Conducty stored state under ~/.conducty/{plans,designs,history,context}/ as flat folders. That worked but produced a write-only log: each file was an island. The vault model fixes this:
- Wikilinks make the relationships first-class. A plan links the designs it consumed, the context files it pulled from, and the improvement experiments it's testing.
- Backlinks are free. Open
Context My-Appand Obsidian shows every plan and design that referenced it. - Graph view turns the orchestration history into a navigable knowledge map.
- Future plans read the past by following links, not by re-grepping flat files.
The vault is the memory of the orchestrator. Treat every note as a node in a graph, not a row in a log.
Layout
Nested by category, flat-within-category. Subfolders organize by note type and scope; basenames stay unique and descriptive so wikilinks (which resolve by basename across all subfolders) are unaffected by directory placement.
Title Case filenames. Wikilinks resolve by basename, so the filename is the canonical title — never rely on the directory path to disambiguate a wikilink target.
Why nested: flat root broke at scale — a vault with 50+ plans + multi-project context becomes a wall of files. Nesting by category makes Obsidian's file tree navigable. Wikilinks are unaffected because Obsidian resolves them by basename across the whole vault, not by path.
Why long unique basenames inside `Context/{Project}/` (e.g. Context My App Architecture.md, not just Architecture.md): wikilinks like [[Architecture]] would collide across projects. Keeping Context {Project} {Slice} as the basename makes every wikilink unambiguous forever.
Two note categories:
- Per-instance notes (timestamped or per-project) — get an index
- Accumulating notes (singular files appended to over time) — no index, they self-index
Naming Conventions
Per-instance notes
| Note type | Directory | Filename pattern | Example | |
|---|---|---|---|---|
| Plan | Plans/ | Plan YYYY-MM-DD HHmm [{Topic Title Case}].md | Plans/Plan 2026-04-27 0930 Auth Cleanup.md | |
| Design | Designs/ | Design YYYY-MM-DD HHmm {Topic Title Case}.md | Designs/Design 2026-04-27 0930 Auth Cleanup.md | |
| Improvement | Improvements/ | Improvement YYYY-MM-DD HHmm.md | Improvements/Improvement 2026-04-27 1830.md | |
| Code review | Code Reviews/ | Code Review YYYY-MM-DD HHmm.md | Code Reviews/Code Review 2026-04-27 1900.md | |
| Ship report | Ship Reports/ | Ship Report YYYY-MM-DD HHmm.md | Ship Reports/Ship Report 2026-04-27 1915.md | |
| Index | Indexes/ | {Topic} Index.md | Indexes/Plans Index.md |
Per-project notes (the context sub-graph — see [[conducty-context]])
All per-project notes live under Context/{Project}/. Optional bounded-context deep notes go under Context/{Project}/Modules/. Refresh delta notes go under Context/{Project}/Refreshes/.
| Note | Path | Purpose | |
|---|---|---|---|
| Hub | Context/{Project}/Context {Project}.md | Project identity, links to slices, last-refreshed | |
| Architecture | Context/{Project}/Context {Project} Architecture.md | Bounded contexts, module map, Mermaid dep graph, seams | |
| Conventions | Context/{Project}/Context {Project} Conventions.md | Naming, style, lint/format commands | |
| Invariants | Context/{Project}/Context {Project} Invariants.md | Public APIs, schemas, contracts | |
| Hotspots | Context/{Project}/Context {Project} Hotspots.md | Frequently changed files, recent activity | |
| Tests | Context/{Project}/Context {Project} Tests.md | Test/typecheck/lint/vuln commands; characterization data. Read by [[conducty-ship]]. | |
| Glossary | Context/{Project}/Context {Project} Glossary.md | Domain terms, ubiquitous language | |
| Module deep note | Context/{Project}/Modules/Context {Project} {Module}.md | Optional, for non-trivial bounded contexts | |
| Refresh delta | Context/{Project}/Refreshes/Context Refresh {Project} YYYY-MM-DD HHmm.md | Diff vs prior refresh — temporal record |
Multiple plans per day are expected — that's why plans (and the artifacts derived from them) carry an HHmm timestamp. The topic suffix on plans is optional but encouraged when more than one plan runs in a day, so the filename alone tells you which plan is which.
Accumulating notes
All accumulating notes live under Accumulators/.
| Note type | Path | Behavior | |
|---|---|---|---|
| Failure Patterns | Accumulators/Failure Patterns.md | Append-prepend: newest pattern at top | |
| Metrics | Accumulators/Metrics.md | Append-prepend: one row per plan run, newest at top | |
| Prompt Log | Accumulators/Prompt Log.md | Append-prepend: one entry per prompt outcome, newest at top |
Accumulating notes never get a timestamp in their filename — they hold every entry across all plan runs and serve as the cross-time learning corpus.
Always Title Case. Always full ISO date for any dated note.
Frontmatter
Every Conducty note carries frontmatter. Minimum required fields:
Add type-specific fields as the skill requires (e.g., plans add appetite, metrics rows add pass_rate).
Linking
Every per-instance note links its peers and at least one index. A note with no outgoing links has been written into a void.
Standard link patterns:
- A plan links:
[[Plans Index]], the design(s) it decomposed ([[Design 2026-04-27 0930 Auth Cleanup]]), the context note(s) it loaded ([[Context My App]]), the prior improvement entry whose experiments it's testing ([[Improvement 2026-04-26 1830]]), and any prior plan whose work carries forward ([[Plan 2026-04-26 0930]]). Plans also link the accumulating notes they will append to:[[Failure Patterns]],[[Metrics]],[[Prompt Log]]. - A design links:
[[Designs Index]], the plan(s) that consume it, the context note(s) for relevant projects, related dialectic decisions. - A context note links:
[[Context Index]]and any plan/design that depends on the project. Backlinks accumulate naturally as plans reference it. - An improvement entry links:
[[Improvements Index]], the plan it reflects on,[[Failure Patterns]], prior improvements being evaluated. - Failure Patterns (accumulating) — each entry inside the file links the plan that surfaced it (
[[Plan 2026-04-27 0930 Auth Cleanup]]) and the prompt entry in[[Prompt Log]]. - Metrics (accumulating) — each row links its plan.
- Prompt Log (accumulating) — each entry links its plan and, if a failure pattern was extracted, the entry in
[[Failure Patterns]].
Place links in two zones:
- Inline — wherever a name appears in prose, wikilink it on first reference within a section.
- Related section at the bottom of per-instance notes:
Index Notes
Index notes are flat lists with frontmatter. They serve two purposes:
- Make Obsidian's left-sidebar useful (navigate by index, not file tree)
- Catch orphaned per-instance notes — if a new note isn't in its index, it'll get lost
Each index follows this shape:
When you create a new per-instance note, always prepend its wikilink to the matching index in the same operation.
Conducty Index is the root and links every index plus the accumulating notes:
Bootstrap
If the vault root doesn't exist, create it (with the standard subdirectories) before any other note write:
Context/{Project}/ and its Modules/ + Refreshes/ subfolders are created lazily by [[conducty-context]] when ingesting a project.
Seed these notes if missing:
This is the start of the file. Read the full SKILL.md on GitHub.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/conducty-obsidian/SKILL.md
Popularity
- Repository stars
- 213
- Repository forks
- 17
- 7-day trend score
- 27
Stars and forks are for the whole GitHub repository, not this skill alone. Refreshed 15 May 2026, 5 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/robertbarclayy/conducty/conducty-obsidian)